Improving the life chances of disadvantaged youth is a top policy priority for the U.S., yet remarkably few effective intervention strategies have been identified - particularly for males. This lack of understanding of how to effectively intervene hels explain why the high school graduation rate in the US today is not much different from in 1970 (despite growing returns to schooling), and why the homicide rate in the US today is not much different from that of 1950 (or 1900). The core theme of the present project (which is a revised version of our previous submission 1-P01-HD076816-01), is to learn more about the most effective (and cost-effective) ways to improve schooling and other long-term life outcomes and reduce risk of violence involvement and delinquency of disadvantaged youth. We address these challenges here with three interrelated projects that would be carried out by our inter-disciplinar research team [including economists Jens Ludwig (PD), Philip Cook, Jonathan Guryan and Roland Fryer, sociologists Susan Mayer, George Farkas and Gretchen Cusick, psychologists Kenneth Dodge and Laurence Steinberg, and public health experts Roseanna Ander and Harold Pollack] in collaboration with the Chicago Public Schools (CPS) and two non-profits service providers (Youth Guidance and Match Education): Projects 1 and 2 involve independent randomizations of the same sample of 2,134 disadvantaged male youth in 9th and 10th grade in 12 CPS high schools to a best-practice academic intervention (intensive math remediation delivered by Match, which seems very promising based on the difference-in-difference study of Fryer, 2011, but has not yet been subject to a large-scale randomized controlled trial, or RCT) and a best-practice non-academic intervention (an adapted version of cognitive behavioral therapy, or CBT, called Becoming a Man for which we have encouraging results from three previous pilots, but no data yet on mechanisms of action). These two projects together create a 2x2 factorial design that lets us test the relative effectiveness (and cost-effectiveness) of best-practice academic vs. non-academic supports for adolescents, and whether intervening in both domains simultaneously yields synergistic (more than additive) effects. We will be able to measure outcomes over time using longitudinal administrative records on schooling, arrests, and labor market outcomes. Project 3 proposes in-person data collection on this same sample to measure outcomes and candidate mediators, including several novel measures of candidate mediators such as schooling expectations (we think key for Match) and meta-cognitive processes (which we think is the likely to be the key mechanism behind BAM effects - and so distinguishes our conceptual framework for this intervention from the social-emotion learning literature). In addition to the administrative core that would help manage and coordinate these interrelated projects, we have strengthened the current revision with two new cores: an implementation and administrative data core, which will among other things collect detailed process measures for use in our analysis, as well as an analysis and dissemination core.